Description
Spark's StringUtils.escapeLikeRegex() method is written inefficiently, performing tons of object allocations due to the use zip(), flatMap() , and mkString. Instead, I think method should be rewritten in an imperative style using a Java string builder.
This method can become a performance bottleneck in cases where regex expressions are used with non-constant-foldable expressions (e.g. the regex expression comes from the data rather than being part of the query).
Here's the code in question: https://github.com/apache/spark/blob/d785217b791882e075ad537852d49d78fc1ca31b/sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/util/StringUtils.scala#L28
Attachments
Issue Links
- is superceded by
-
SPARK-17647 SQL LIKE does not handle backslashes correctly
-
- Resolved
-
- links to
